ai 6.0.0-beta.45 → 6.0.0-beta.47

package/CHANGELOG.md

@@ -1,5 +1,18 @@
 # ai
 
+## 6.0.0-beta.47
+
+### Patch Changes
+
+- c62ecf0: feat(ai): add support for v2 specs in transcription and speech models
+
+## 6.0.0-beta.46
+
+### Patch Changes
+
+- Updated dependencies [0a2ff8a]
+  - @ai-sdk/gateway@2.0.0-beta.30
+
 ## 6.0.0-beta.45
 
 ### Patch Changes

package/dist/index.d.mts

@@ -3,7 +3,7 @@ import * as _ai_sdk_provider_utils from '@ai-sdk/provider-utils';
 import { Tool, InferToolInput, InferToolOutput, AssistantModelMessage, ToolModelMessage, ReasoningPart, ModelMessage, FlexibleSchema, InferSchema, SystemModelMessage, UserModelMessage, ProviderOptions, IdGenerator, ToolCall, DataContent, Resolvable, FetchFunction } from '@ai-sdk/provider-utils';
 export { AssistantContent, AssistantModelMessage, DataContent, FilePart, FlexibleSchema, IdGenerator, ImagePart, InferSchema, InferToolInput, InferToolOutput, ModelMessage, Schema, SystemModelMessage, TextPart, Tool, ToolApprovalRequest, ToolApprovalResponse, ToolCallOptions, ToolCallPart, ToolContent, ToolExecuteFunction, ToolModelMessage, ToolResultPart, UserContent, UserModelMessage, asSchema, createIdGenerator, dynamicTool, generateId, jsonSchema, parseJsonEventStream, tool, zodSchema } from '@ai-sdk/provider-utils';
 import * as _ai_sdk_provider from '@ai-sdk/provider';
-import { EmbeddingModelV3, EmbeddingModelV2, EmbeddingModelV3Embedding, ImageModelV3, ImageModelV3CallWarning, ImageModelV3ProviderMetadata, JSONValue as JSONValue$1, LanguageModelV3, LanguageModelV2, LanguageModelV3FinishReason, LanguageModelV3CallWarning, LanguageModelV3Source, LanguageModelV3Middleware, SharedV3ProviderMetadata, SpeechModelV3, SpeechModelV3CallWarning, TranscriptionModelV3, TranscriptionModelV3CallWarning, LanguageModelV3Usage, LanguageModelV3CallOptions, AISDKError, LanguageModelV3ToolCall, JSONSchema7, JSONParseError, TypeValidationError, ProviderV3, ProviderV2, NoSuchModelError, JSONObject } from '@ai-sdk/provider';
+import { EmbeddingModelV3, EmbeddingModelV2, EmbeddingModelV3Embedding, ImageModelV3, ImageModelV3CallWarning, ImageModelV3ProviderMetadata, JSONValue as JSONValue$1, LanguageModelV3, LanguageModelV2, LanguageModelV3FinishReason, LanguageModelV3CallWarning, LanguageModelV3Source, LanguageModelV3Middleware, SharedV3ProviderMetadata, SpeechModelV3, SpeechModelV2, SpeechModelV3CallWarning, TranscriptionModelV3, TranscriptionModelV2, TranscriptionModelV3CallWarning, LanguageModelV3Usage, LanguageModelV3CallOptions, AISDKError, LanguageModelV3ToolCall, JSONSchema7, JSONParseError, TypeValidationError, ProviderV3, ProviderV2, NoSuchModelError, JSONObject } from '@ai-sdk/provider';
 export { AISDKError, APICallError, EmptyResponseBodyError, InvalidPromptError, InvalidResponseDataError, JSONParseError, JSONSchema7, LoadAPIKeyError, NoContentGeneratedError, NoSuchModelError, TooManyEmbeddingValuesForCallError, TypeValidationError, UnsupportedFunctionalityError } from '@ai-sdk/provider';
 import { ServerResponse } from 'node:http';
 import { AttributeValue, Tracer } from '@opentelemetry/api';
@@ -164,7 +164,7 @@ type ProviderMetadata = SharedV3ProviderMetadata;
 /**
 Speech model that is used by the AI SDK Core functions.
   */
-type SpeechModel = SpeechModelV3;
+type SpeechModel = string | SpeechModelV3 | SpeechModelV2;
 /**
 Warning from the model provider for this call. The call will proceed, but e.g.
 some settings might not be supported, which can lead to suboptimal results.
@@ -193,7 +193,7 @@ type SpeechModelResponseMetadata = {
 /**
 Transcription model that is used by the AI SDK Core functions.
   */
-type TranscriptionModel = TranscriptionModelV3;
+type TranscriptionModel = string | TranscriptionModelV3 | TranscriptionModelV2;
 /**
 Warning from the model provider for this call. The call will proceed, but e.g.
 some settings might not be supported, which can lead to suboptimal results.
@@ -3691,7 +3691,7 @@ declare function generateSpeech({ model, text, voice, outputFormat, instructions
     /**
   The speech model to use.
        */
-    model: SpeechModelV3;
+    model: SpeechModel;
     /**
   The text to convert to speech.
      */
@@ -4149,7 +4149,7 @@ declare function transcribe({ model, audio, providerOptions, maxRetries: maxRetr
     /**
   The transcription model to use.
        */
-    model: TranscriptionModelV3;
+    model: TranscriptionModel;
     /**
   The audio data to transcribe.
      */

package/dist/index.d.ts

@@ -3,7 +3,7 @@ import * as _ai_sdk_provider_utils from '@ai-sdk/provider-utils';
 import { Tool, InferToolInput, InferToolOutput, AssistantModelMessage, ToolModelMessage, ReasoningPart, ModelMessage, FlexibleSchema, InferSchema, SystemModelMessage, UserModelMessage, ProviderOptions, IdGenerator, ToolCall, DataContent, Resolvable, FetchFunction } from '@ai-sdk/provider-utils';
 export { AssistantContent, AssistantModelMessage, DataContent, FilePart, FlexibleSchema, IdGenerator, ImagePart, InferSchema, InferToolInput, InferToolOutput, ModelMessage, Schema, SystemModelMessage, TextPart, Tool, ToolApprovalRequest, ToolApprovalResponse, ToolCallOptions, ToolCallPart, ToolContent, ToolExecuteFunction, ToolModelMessage, ToolResultPart, UserContent, UserModelMessage, asSchema, createIdGenerator, dynamicTool, generateId, jsonSchema, parseJsonEventStream, tool, zodSchema } from '@ai-sdk/provider-utils';
 import * as _ai_sdk_provider from '@ai-sdk/provider';
-import { EmbeddingModelV3, EmbeddingModelV2, EmbeddingModelV3Embedding, ImageModelV3, ImageModelV3CallWarning, ImageModelV3ProviderMetadata, JSONValue as JSONValue$1, LanguageModelV3, LanguageModelV2, LanguageModelV3FinishReason, LanguageModelV3CallWarning, LanguageModelV3Source, LanguageModelV3Middleware, SharedV3ProviderMetadata, SpeechModelV3, SpeechModelV3CallWarning, TranscriptionModelV3, TranscriptionModelV3CallWarning, LanguageModelV3Usage, LanguageModelV3CallOptions, AISDKError, LanguageModelV3ToolCall, JSONSchema7, JSONParseError, TypeValidationError, ProviderV3, ProviderV2, NoSuchModelError, JSONObject } from '@ai-sdk/provider';
+import { EmbeddingModelV3, EmbeddingModelV2, EmbeddingModelV3Embedding, ImageModelV3, ImageModelV3CallWarning, ImageModelV3ProviderMetadata, JSONValue as JSONValue$1, LanguageModelV3, LanguageModelV2, LanguageModelV3FinishReason, LanguageModelV3CallWarning, LanguageModelV3Source, LanguageModelV3Middleware, SharedV3ProviderMetadata, SpeechModelV3, SpeechModelV2, SpeechModelV3CallWarning, TranscriptionModelV3, TranscriptionModelV2, TranscriptionModelV3CallWarning, LanguageModelV3Usage, LanguageModelV3CallOptions, AISDKError, LanguageModelV3ToolCall, JSONSchema7, JSONParseError, TypeValidationError, ProviderV3, ProviderV2, NoSuchModelError, JSONObject } from '@ai-sdk/provider';
 export { AISDKError, APICallError, EmptyResponseBodyError, InvalidPromptError, InvalidResponseDataError, JSONParseError, JSONSchema7, LoadAPIKeyError, NoContentGeneratedError, NoSuchModelError, TooManyEmbeddingValuesForCallError, TypeValidationError, UnsupportedFunctionalityError } from '@ai-sdk/provider';
 import { ServerResponse } from 'node:http';
 import { AttributeValue, Tracer } from '@opentelemetry/api';
@@ -164,7 +164,7 @@ type ProviderMetadata = SharedV3ProviderMetadata;
 /**
 Speech model that is used by the AI SDK Core functions.
   */
-type SpeechModel = SpeechModelV3;
+type SpeechModel = string | SpeechModelV3 | SpeechModelV2;
 /**
 Warning from the model provider for this call. The call will proceed, but e.g.
 some settings might not be supported, which can lead to suboptimal results.
@@ -193,7 +193,7 @@ type SpeechModelResponseMetadata = {
 /**
 Transcription model that is used by the AI SDK Core functions.
   */
-type TranscriptionModel = TranscriptionModelV3;
+type TranscriptionModel = string | TranscriptionModelV3 | TranscriptionModelV2;
 /**
 Warning from the model provider for this call. The call will proceed, but e.g.
 some settings might not be supported, which can lead to suboptimal results.
@@ -3691,7 +3691,7 @@ declare function generateSpeech({ model, text, voice, outputFormat, instructions
     /**
   The speech model to use.
        */
-    model: SpeechModelV3;
+    model: SpeechModel;
     /**
   The text to convert to speech.
      */
@@ -4149,7 +4149,7 @@ declare function transcribe({ model, audio, providerOptions, maxRetries: maxRetr
     /**
   The transcription model to use.
        */
-    model: TranscriptionModelV3;
+    model: TranscriptionModel;
     /**
   The audio data to transcribe.
      */

package/dist/index.js

@@ -581,6 +581,24 @@ function transformToV3EmbeddingModel(model) {
     }
   });
 }
+function transformToV3TranscriptionModel(model) {
+  return new Proxy(model, {
+    get(target, prop) {
+      if (prop === "specificationVersion")
+        return "v3";
+      return target[prop];
+    }
+  });
+}
+function transformToV3SpeechModel(model) {
+  return new Proxy(model, {
+    get(target, prop) {
+      if (prop === "specificationVersion")
+        return "v3";
+      return target[prop];
+    }
+  });
+}
 function resolveLanguageModel(model) {
   if (typeof model !== "string") {
     if (model.specificationVersion !== "v3" && model.specificationVersion !== "v2") {
@@ -617,6 +635,42 @@ function resolveEmbeddingModel(model) {
     model
   );
 }
+function resolveTranscriptionModel(model) {
+  var _a17, _b;
+  if (typeof model !== "string") {
+    if (model.specificationVersion !== "v3" && model.specificationVersion !== "v2") {
+      const unsupportedModel = model;
+      throw new UnsupportedModelVersionError({
+        version: unsupportedModel.specificationVersion,
+        provider: unsupportedModel.provider,
+        modelId: unsupportedModel.modelId
+      });
+    }
+    if (model.specificationVersion === "v2") {
+      return transformToV3TranscriptionModel(model);
+    }
+    return model;
+  }
+  return (_b = (_a17 = getGlobalProvider()).transcriptionModel) == null ? void 0 : _b.call(_a17, model);
+}
+function resolveSpeechModel(model) {
+  var _a17, _b;
+  if (typeof model !== "string") {
+    if (model.specificationVersion !== "v3" && model.specificationVersion !== "v2") {
+      const unsupportedModel = model;
+      throw new UnsupportedModelVersionError({
+        version: unsupportedModel.specificationVersion,
+        provider: unsupportedModel.provider,
+        modelId: unsupportedModel.modelId
+      });
+    }
+    if (model.specificationVersion === "v2") {
+      return transformToV3SpeechModel(model);
+    }
+    return model;
+  }
+  return (_b = (_a17 = getGlobalProvider()).speechModel) == null ? void 0 : _b.call(_a17, model);
+}
 function getGlobalProvider() {
   var _a17;
   return (_a17 = globalThis.AI_SDK_DEFAULT_PROVIDER) != null ? _a17 : import_gateway.gateway;
@@ -814,7 +868,7 @@ function detectMediaType({
 var import_provider_utils2 = require("@ai-sdk/provider-utils");
 
 // src/version.ts
-var VERSION = true ? "6.0.0-beta.45" : "0.0.0-test";
+var VERSION = true ? "6.0.0-beta.47" : "0.0.0-test";
 
 // src/util/download/download.ts
 var download = async ({ url }) => {
@@ -8327,12 +8381,9 @@ async function generateSpeech({
   headers
 }) {
   var _a17;
-  if (model.specificationVersion !== "v3") {
-    throw new UnsupportedModelVersionError({
-      version: model.specificationVersion,
-      provider: model.provider,
-      modelId: model.modelId
-    });
+  const resolvedModel = resolveSpeechModel(model);
+  if (!resolvedModel) {
+    throw new Error("Model could not be resolved");
   }
   const headersWithUserAgent = (0, import_provider_utils25.withUserAgentSuffix)(
     headers != null ? headers : {},
@@ -8343,7 +8394,7 @@ async function generateSpeech({
     abortSignal
   });
   const result = await retry(
-    () => model.doGenerate({
+    () => resolvedModel.doGenerate({
       text: text2,
       voice,
       outputFormat,
@@ -9746,12 +9797,9 @@ async function transcribe({
   abortSignal,
   headers
 }) {
-  if (model.specificationVersion !== "v3") {
-    throw new UnsupportedModelVersionError({
-      version: model.specificationVersion,
-      provider: model.provider,
-      modelId: model.modelId
-    });
+  const resolvedModel = resolveTranscriptionModel(model);
+  if (!resolvedModel) {
+    throw new Error("Model could not be resolved");
   }
   const { retry } = prepareRetries({
     maxRetries: maxRetriesArg,
@@ -9765,7 +9813,7 @@ async function transcribe({
   const result = await retry(
     () => {
       var _a17;
-      return model.doGenerate({
+      return resolvedModel.doGenerate({
         audio: audioData,
         abortSignal,
         headers: headersWithUserAgent,